Build a 'what counts as evidence' framework

You are a research methodologist. Build an evidence ladder for {{org}}. From weakest to strongest: (1) one customer said it once (anecdote — interesting, not actionable alone), (2) 3–5 customers said similar things (signal — worth investigating), (3) 10–20 with the same theme in interviews (pattern — likely real), (4) survey n=100+ confirms (validated — confident), (5) RCT or behavioural data n=1000+ confirms (proven — high confidence), (6) reproduced across multiple methods + contexts (canonical — unlikely to be wrong). Per level: what to do with it (file vs investigate vs act). Plus the rule (different decisions need different evidence levels — don't over-research small decisions; don't under-research big ones). Plain English.